The PC Just Got Its Most Consequential Redesign Since Windows 95
Jensen Huang took the stage at Computex Taipei on June 1, 2026 and made a claim that few technology executives would risk: that the personal computer is being reinvented, and that the reinvention is as significant as the transformation of the phone into the smartphone. "This reinvention of the computer is as big of a deal as the reinvention of the phone into what we now know as the smartphone," Huang told delegates as he unveiled the RTX Spark — a superchip combining an Arm CPU, a Blackwell GPU, and 128 gigabytes of unified memory onto a single die, designed specifically to run AI agents locally on Windows laptops and compact desktop PCs. Nvidia on Monday unveiled the new powerful chips that would bring advanced artificial intelligence functions into laptops and desktop computers, with new PC models from brands including Microsoft Surface, Dell, HP, Asus, Lenovo, MSI, Acer, and Gigabyte set to roll out later this year. Nvidia shares climbed more than 4% in early trading. Intel fell more than 6%. AMD dropped over 3%. The market rendered its verdict on who wins a PC architecture transition away from x86 in a single session.
The RTX Spark brings together 30 years of Nvidia innovation — including CUDA, RTX, DLSS, FP4, TensorRT, OptiX, Reflex, and G-SYNC — to slim Windows laptops and compact ultra-efficient desktop PCs. The 128GB of unified memory is the specification that defines RTX Spark's AI capability: running a frontier AI model as a local agent requires loading model weights into memory, and most current laptops have 16 to 32GB of RAM. At 128GB, the RTX Spark has enough memory to run models equivalent in capability to GPT-4 class locally, without cloud connectivity, without latency, and without the privacy concern of sending prompts to external servers. When Huang described the new user experience — "you could talk to it, it could look at you, you could ask it to read files, go help you do some research, it could do a lot more" — he was not describing a smarter Siri. He was describing a persistent AI agent that understands the user's context, has access to their files, can take actions across applications, and can complete multi-step tasks without a prompt for each step.
What RTX Spark Does to Intel and the x86 Architecture
Intel's 6% single-day decline on Nvidia's RTX Spark announcement is the market's assessment of a competitive threat that Intel's own roadmap cannot neutralise in the timeframe that matters. Intel has dominated the PC processor market since the IBM PC's 1981 debut — a 45-year hold on the world's most popular computing platform. The RTX Spark does not attack Intel on Intel's terms. It offers a different architecture entirely, purpose-built for a different mode of computing. Intel's answer to an AI agent running locally on 128GB of unified memory is a laptop CPU that connects to external RAM chips over a bus. The latency, bandwidth, and power efficiency disadvantages of that architecture in an AI workload context are structural — unfixable with a faster clock speed or a more advanced manufacturing process. Nvidia's commitment to future Vera Rubin Spark and Rosa Feynman Spark generations makes this a platform strategy rather than a product experiment. The autumn 2026 launch from eight major PC brands simultaneously — without precedent in Windows hardware history — signals that the industry has already chosen its architecture for the AI PC era.
The CUDA Advantage: Why the Software Moat Extends to Every Desk
The CUDA compatibility of RTX Spark is its most durable competitive advantage. The entire ecosystem of AI software built on CUDA for data centres runs on RTX Spark devices without modification. Developers who have built AI applications for Nvidia's server GPUs get a personal computing target for free. The software moat that has been Nvidia's most defensible position in the data centre — where switching away from CUDA requires rewriting code, retraining engineering teams, and accepting performance uncertainty — extends to the PC the moment RTX Spark ships. Apple proved with its M-series chips that a unified memory architecture dramatically outperforms conventional CPU-plus-discrete-GPU designs for AI workloads. Nvidia is applying the same architectural principle to Windows, but with the additional advantage of CUDA compatibility that Apple's platform, for all its performance merits, cannot offer developers building cross-platform AI applications. The PC installed base of approximately 1.4 billion devices represents the largest single hardware upgrade opportunity Nvidia has ever addressed — and RTX Spark is the product designed to capture it.